The Andersen P2P file sharing study on the purchase of music CDs in Canada

In 2006, the highly regarded economics professor Prof. Liebowitz, Director of the Center for Economic Analysis of Property Rights and Innovation at University of Texas, surveyed the entire field of econometric studies on file sharing. On the basis of his comprehensive review (which displayed a remarkable consensus on the issue), he concluded that “file-sharing has brought significant harm to the recording industry”. Prior to that in a comprehensive article published in 2005 Prof. Liebowitz criticized the theory that unlicensed file sharing helps copyright owners. He said those that professed this view saw “gains from copying in every nook and cranny of the economy, when in reality the instances of such gains are likely to be rather limited.”

Despite the exhaustive work of Prof. Liebowitz and others, Industry Canada retained Birgitte Andersen, a Danish academic working as a Reader in London, England, at Birkbeck College, to study the impact of P2P file-sharing on the purchase of music in Canada. Her report, titled The Impact of Music Downloads and P2P File-Sharing on the Purchase of Music:  A Study for Industry Canada, was published in 2007 on Industry Canada’s website. It confirmed that a “review of existing econometric studies suggests that P2P file-sharing tends to decrease music purchasing.” However, based on a study of evidence from new consumer surveys, the study counter-intuitively concluded “the opposite, namely that P2P file sharing tends to increase rather than decrease music purchasing.”

Immediately after the study was published it was acclaimed as a “must read” by Michael Geist. He used the study to further his copyright agenda claiming that the music industry “has benefited from P2P” and “that there is no “emergency” that necessitates legislative intervention” to reform the Copyright Act. Over the years, Michael Geist continued to cite the study and to defend Ms. Andersen, including when her paper was criticized by Prof. Liebowitz, who after analyzing the study found “a result that is not only implausible but is actually impossible to be true”. Michael Geist vouched for the study, claiming it was “not a study with a particular desired outcome”. Yet at the time she was retained to do the study, Ms. Andersen’s published writings evidenced an antagonism towards both the music industry and the copyright system.[1]

In March 2010, Ms. Andersen backed away significantly from the published findings in her 2007 study. In a revised paper she dropped her claim that unauthorized file sharing helped the music industry. However, along with her co-author Marion Frenz, she now claimed instead that “the Industry Canada data showed no association between the number of P2P files downloaded and CD album sales”.  She concluded that “P2P file-sharing is not to blame for the decline in CD markets”.

Despite Ms Andersen’s own abandonment of the key conclusion in the 2007 study, Industry Canada continues inexplicably to post the abandoned 2007 study on its website or to even publish a notice that the conclusions in the study were abandoned by its authors, who published a subsequent paper using the same data that came to a different conclusion.

Earlier this week, Prof. George Barker, the Director of the Centre for Law and Economics, at the College Of Law, Australian National University, published a comprehensive study using the same raw data used by Ms. Andersen to re-examine the findings in her two papers. Prof. Barker subjected the data to a rigorous analysis, correcting for what he identified as serious flaws in Ms. Andersen’s studies. He found that not only did the underlying data not support the conclusions in either of the previous studies, it supported the exact opposite conclusion.

Prof. Barker’s conclusion was that both the initial and the reworked Andersen studies were fundamentally flawed. When the same raw data was analyzed correcting for the errors he identified, it showed that P2P file-sharing reduced demand for CDs. In particular, that “a 10% increase in P2P downloads reduces CD demand by around 0.4%”.

Prof. Barker’s summarized his corrected analysis of the Industry Canada data as follows:

All of our regression results (reported in Tables 2 and 3 above) show a negative association between P2P downloading and CD demand.  We consistently find a negative and statistically significant partial correlation between CD purchases and P2P downloads.  The range of these estimated correlations is between -0.039 and -0.050, with more consistent, mid-range values coming from the difference on difference regressions (-0.041 and -0.043).  This implies a 10% increase in P2P downloads reduces CD demand by around 0.4%.

This directly contradicts the much cited and controversial conclusion of Andersen and Frenz… AF claimed that the data showed “… no association between the number of P2P files downloaded and CD album sales,” claiming therefore that “… this paper show (sic) that P2P file-sharing is not to blame for the decline in CD markets.  Music markets are not simply undermined by free music downloading and P2P file-sharing.”

In this paper we have corrected for two fundamental errors in the previous analysis by AF leading to their erroneous conclusion. First we corrected for the fact AF biased their results by excluding from their analysis the group of consumers who had completely stopped purchasing CD’s (potentially because of P2P activity) prior to 2005.  This is the very group who were most responsive, or likely to have substituted P2P downloading for CD purchases.  Second we controlled for the fact that the level of an individual’s demand for CD’s, and the level of an individual’s P2P downloading may be correlated simply because they are both affected by the same third factor, such as love of music, so that high (or low) levels of CD demand is likely to be associated with high (low) levels of P2P demand.  Such a positive association between the level of demand and level of P2P downloading may have led AF to mistakenly conclude they had found evidence of a positive market creation effect, as AF regressed the level of individuals CD demand against the level of individuals P2P downloading.  Instead we focused on the changes in CD demand and changes in P2P downloading, using data in the survey that AF ignored on 2004 and 2005 behaviours of participants.  By focusing on a longitudinal analysis of how the change in individual P2P downloading behaviour affected the change in CD demand we were better able to isolate the true relationship between increases or changes in P2P downloading and changes in CD demand.

Prof. Barker’s conclusions about the Industry Canada data are consistent with what Prof. Liebowitz concluded before the first Andersen paper was published, namely, that the vast majority of econometric studies conducted around the world show that unauthorized peer-to-peer file sharing of music harms sales of music.

All the same, the evidence here supports the current findings from almost all econometric studies that have been undertaken to date, including those in this issue—file sharing has brought significant harm to the recording industry.  The birth of file sharing and the very large decline in CD sales that immediately followed is a powerful piece of evidence on its own.  The 2004 increase in CD sales, temporarily reversing the decline, largely matches a reversal in the amount of file-sharing activity.  Furthermore, analysis of the various possible alternative explanations for the decline in CD sales fails to find any viable candidates.

This conclusion, preliminary though it might be, should not be much of a surprise.  Common sense is, or should be, the handmaiden of economic analysis.  When given the choice of free and convenient high-quality copies versus purchased originals, is it really a surprise that a significant number of individuals will choose to substitute the free copy for the purchase?  The conditions needed to override this basic intuition are demanding and seemingly not met in the case of file sharing.

Prof. Barker’s study once again raises fundamental questions about the Andersen study. One could ask why Industry Canada commissioned the study, why the department selected Ms. Andersen to be the primary author of it, how it chose to vet the study`s conclusions before publication, whether it had a desired outcome, and why it was published when it was. These are questions to which we may never know the complete answer.[2]

What we do know is that Industry Canada continues to publish the original 2007 Andersen study on its website. It does so even after the conclusions were abandoned by its authors and after distinguished economists have detailed fundamental reasons for doubting its correctness.

The 2007 Andersen study is available in PDF form from Industry Canada’s website through a direct link to a page hosted by Industry Canada. Industry Canada also publishes a webpage version on the Industry Canada Intellectual Property Directorate site. Both versions contain a disclaimer that the views in the paper are not to be attributed to Industry Canada. In the PDF version of the paper, the disclaimer appears on the first page and states: “THE VIEWS EXPRESSED ARE THOSE OF THE AUTHORS. NO RESPONSIBILITY FOR THEM SHOULD BE ATTRIBUTED TO INDUSTRY CANADA.”

The Industry Canada webpage version contains a somewhat different disclaimer that is published by the Intellectual Property Directorate of Industry Canada. This disclaimer says:

These studies were funded by the Intellectual Property Policy Directorate, Industry Canada. The opinions expressed in these studies are those of the respective authors and do not necessarily reflect the policies or opinions of Industry Canada or the Government of Canada.

This Industry Canada disclaimer is accessible only through a link at the bottom of the page to the word “disclaimer”. Readers will not even be aware of it unless they scroll down to the bottom of the page and click on the link. Like a browse wrap (web wrap) agreement that courts don’t enforce when the link to the agreement is buried at the bottom of a web page, the Industry Canada link to the disclaimer is hardly sufficient notice to readers that Industry Canada does not endorse the opinions in the study.[3]

As the paper continues to be published by Industry Canada, readers, who may be excused for not researching the economics literature to assess its accuracy or lineage, will undoubtedly wrongly conclude that Industry Canada does, in fact, endorse its contents. Moreover, the study has been referred to by third parties such as Michael Geist as the “Industry Canada Sponsored P2P Study”, the “Industry Canada P2P Study” and one of the “Industry Canada studies on the music industry”. The public reading these blogs, which are still being disseminated, will not likely be aware of the Industry Canada disclaimer or the admitted inaccuracies in, or criticisms of, the study.

The continued dissemination of the study could lead even objective researchers to imply an association or endorsement by Industry Canada. For example, The Library of Parliament in its Legislative Summary of Bill C-11 referred to the Andersen 2007 study as an “Industry Canada released study investigating the effects on music sales of music downloading”.

In light of its acknowledged inaccuracies and the continued potential of the 2007 Andersen study to mislead the public about what the Industry Canada data shows and whether the opinions in the study are endorsed by Industry Canada, the study should be removed from Industry Canada’s website. At the very least, like cigarettes, it should be appropriately labelled with a prominent notice.

* Update: Industry Canada subsequently removed the webpage version of the study from its website.